You are on page 1of 15

Applied Mathematical Sciences, Vol. 6, 2012, no.

28, 1347 - 1361

An Improved Parameter Regula Falsi Method


for Enclosing a Zero of a Function
Norhaliza Abu Bakar1 , Mansor Monsi1 , Nasruddin Hassan2

1
Department of Mathematics, Faculty of Science
Universiti Putra Malaysia
43400 UPM Serdang, Selangor, Malaysia

2
School of Mathematical Sciences, Faculty of Science and Technology
Universiti Kebangsaan Malaysia
43600 UKM Bangi, Selangor, Malaysia
n izaruedge@yahoo.com

Abstract
An improved parameter regula falsi method p(l)-RF based on a di-
rect generalization of the interval parameter regula falsi p-RF method
is proposed in this paper. This method is modified by using the mid-
point of the current interval in the algorithm and the additional inner
iterations l to improve its rate of convergence. This modification on
p-RF method is then verified on several test examples. Based on the
numerical results and CPU time of p(l)-RF method, it is very clear that
this method performs very well compared to the original method.

Mathematics Subject Classification: 65B99, 65G40

Keywords: CPU time, interval analysis, parameter regula falsi, rate of


convergence, zero of a function

1 Introduction
Apart from iterative procedures, the main tool to be used in this paper is inter-
val analysis based upon the very simple idea of enclosing the zero of a function.
Researchers such as [5] and [2] have shown that the iterative procedures which
involved interval analysis approach is much convenience because the nal in-
terval of the iterative procedure always contain the zero of the function. In
other word, the zero is always bounded in the interval.
1348 N. A. Bakar, M. Monsi and N. Hassan

Parameter regula falsi method p-RF is an extended interval version of Reg-


ula Falsi method (RF). The modication of RF method had been done by
various researchers such as [10] and [7] and only one by [9] on p-RF method.
p-RF method has higher-order of convergence even though only one function
value used [4]. We consider the equation f (x) = 0 and the function f has
a simple zero in the initial interval X (0) . The rate of convergence of p-RF
1 
method [4] is at least (p + p2 + 4) (p 2).
2
In the next section, we described briey about p-RF method and then we
present our modied method called p(l)-RF method. We also described the
p(l)-RF algorithm and then we compare the CPU time and the number of
iterations obtained by p(l)-RF method and p-RF method.

2 p-RF Algorithm
The following is the algorithm for p-RF method [4].
(0) (0)    (k+1) 
(k+1) 
Step 0 : Given that: X (0) = x1 , x2 ; X (0) ; d X (k+1) = x2 x1 
with conditions

f  (x) H = [h1 , h2 ] = {f  (x)|h1 f  (x) h2 } , x X ; 0


/H (1)

f  (x) K = [k1 , k2 ] = {f  (x)|k1 f  (x) k2 } x X (2)

Step 1 : Set k = 0, i = 2


Step 2 : x(k) = m  X (k) (midpoint of X (k) )
 (k) 
f x
(k)
Step 3 : X (k+1) = x(k) X
H
Step 4 : k := k + 1 
Step 5 : x(k) = m X (k)   (k) 
f x
(k)
Step 6 : X (k+1,i2) = x(k) X
H
Step 7 : Compute
x(k) x(k1)
X (k+1,i1) = x(k) f (x(k) )+
f (x ) f (x
(k) (k1) ) 
1  (k+1,i2) (k)
  (k+1,i2) (k1)

(k+1,i2)
K X x X x X
2  
if f x(k) = 0. Otherwise, X (k+1,i1) = X (k+1,i2)
 
Step 8 : z (i) = m X (k+1,i1)
Step 9 : Compute
An improved parameter regula falsi method 1349


(k+1,i) z (i) x(k) (i)
X = z (i) f (z )+
f (z ) f (x )
(i) (k)

1  (k+1,i1) (i)
  (k+1,i1) (k)

(k+1,i1)
K X z X x X
2  
if f x(k) = 0. Otherwise, X (k+1,i) = X (k+1,i1)

Step 10 : If i < p then i := i + 1 and go to 8


Step 11 : X (k+1) = X (k+1,p)
(k+1)
Step 12 : Ifthe width of the interval X greater than , that is
(k+1)
d X > , then go to 4
Step 13 : Stop.

Theorem 1 Let the function f be twice be continuously dierentiable in the


interval X and assume that f has a zero in X. Let the conditions (1) and
(2) be satised. The parameter p is given as an integer number for which
p 2. Then, the sequence {X (k) } calculated according to algorithm p-RF
method satises for p 2:

X (k) , k0

X (0) X (1) X (2) ... with lim X (k) =


k

and
   (p)  (k1) 
d X (k+1) d X (k) d X ; ( 0)
     
where d X (k+1) , d X (k) , and d X (k1) are the width of the intervals
X (k+1) , X (k) and X (k1) respectively. Then the R-order of convergence of p-
RF satises the inequality
1  
OR ((p RF) , ) p + p2 + 4 (3)
2
The proof of these theorem is available in [4].

3 The p(l)-RF Algorithm


Now, we present the algorithm of p(l)-RF  method.
(0) (0) (0) (0)
 (k+1)   (k+1) 
(k+1) 
Step 0 : Given that: X = x1 , x2 ; X ; d X = x2 x1 
with conditions (1) and (2)
Step 1 : Set k = 0, i = 2 and l = 2
1350 N. A. Bakar, M. Monsi and N. Hassan

 
Step 2 : x(k) = m X (k) (midpoint of X (k) )
  (k) 
f x
(k)
Step 3 : X (k+1) = x(k) X
H
Step 4 : k := k + 1
 
Step 5 : x(k) = m X (k)
  (k) 
f x
(k)
Step 6 : X (k+1,i2) = x(k) X
H
Step 7 : Compute

x(k) x(k1)
X (k+1,i1)
= x(k) f (x(k) )+
f (x(k) ) f (x(k1) )

1  (k+1,i2) (k)
  (k+1,i2) (k1)

(k+1,i2)
K X x X x X
2
 
if f x(k) = 0. Otherwise, X (k+1,i1) = X (k+1,i2)
   
Step 8 : z (i) = m X (k+1,i1) and z (i1) = m X (k+1,i2)
Step 9 : Compute

z (i) z (i1) (i)
X (k+1,i) = z (i) f (z )+
f (z (i) ) f (z (i1) )

1  (k+1,i1) (i)
  (k+1,i1) (i1)

(k+1,i1)
K X z X z X
2
 
if f z (i) = 0. Otherwise, X (k+1,i) = X (k+1,i1)
 
Step 10 : If d X (k+1,i) < , go to 18
Step 11 : If i < p then i := i + 1 and go to 8
Step 12 : X (k+1) = X (k+1,p) = X (k+1,p,l1)
Step 13 : Compute

(k+1,p,l) (l) z (l) z (l1) (l)
X = z f (z )+
f (z (l) ) f (z (l1) )

1  (k+1,p,l1) (l)
  (k+1,p,l1) (l1)

(k+1,p,l1)
K X z X z X
2
 (l) 
if f z = 0. Otherwise, X (k+1,p,l) = X (k+1,p,l1)
 
Step 14 : If d X (k+1,p,l) < , go to 18
Step 15 : If l < p then l := l + 1 and go to 13
Step 16 : X (k+1) = X (k+1,p,p)
 
Step 17 : If d X (k+1) > , go to 4
Step 18 : Stop.
An improved parameter regula falsi method 1351

4 The Rate of Convergence of p(l)-RF Method


The p(l)-RF method which is based on p-RF method has its own speciality.
In this modication, we keep the computation process updated for every inner
iteration i. Therefore, we compute another midpoint z (i1) to replaced the
value of midpoint x(k) in Step 9 (refer to Section 2).

Not only that, we also introduce another inner iteration, l (l 2). This
iteration process from l = 2, 3, . . . , p takes place after the computation of inner
iteration i nished. Then, we substitute the latest interval that is X (k+1,p)
with X (k+1,p,l). Readers can track down the modication that we have done
by comparing algorithm on Section 2 and Section 3.

The following theorem is very useful in order to determine the rate of


convergence of a procedure I [4].
Theorem 2 Let I be an interval iteration procedure with the limit x and
let (I, x ) be the set of all sequences generated by I having the properties
that limk x(k) = x and x x(k) , k 0. If there exist a p 1 and a
constant such that for all {x(k) } (I, x ) and for a norm . it holds that
x(k+1)  x(k) p . Then, it follows that the R-order of convergence of I
satises the inequality
OR (I, x ) p
or the R-order of convergence of I is at least p.
Based on the Theorem 2, then we have the following theorem for p(l)-RF
method.
Theorem 3 Let the function f be twice continuously dierentiable in the in-
terval X and assume that f only has a simple real zero in X. Furthermore,
we have intervals H and K satisfying the conditions (1) and (2). Then, the
sequence calculated from the p(l)-RF method satises (p 2):
X (k) , k0 (4)

X (0) X (1) X (2) (5)

lim X (k) = (6)


k

Then, the R-order of convergence of the p(l)-RF method is at least


  
1 2
(2p + (l 2)) + (2p + (l 2)) + 4 (7)
2
1352 N. A. Bakar, M. Monsi and N. Hassan

where (p 2) ; (1 l p) or
  
1 2
OR ((p(l) RF ) , ) (2p + (l 2)) + (2p + (l 2)) + 4 .
2
Proof. :
Of (4), (5) and (6) : Can be found in [4].
Of (7) : The proof of
 
 (k+1,1)  1  K   (k)   (k1) 
d X   d X d X ,
4 H
can be found in [4]. Next, we have

 (k+1,2)  (2) z (2) z (1)  (2) 
d X d z f z +
f (z (2) ) f (z (1) )

1  (k+1,1) (2)
  (k+1,1) (1)

K X z X z
2
 
z (2) z (1) 1  (k+1,1) (2)
  (k+1,1) (1)

= d K X z X z .
f (z (2) ) f (z (1) ) 2
z (2) z (1) 1 1
By = (f  ( )) ( X), it follows that
f (z ) f (z )
(2) (1) H
 
 (k+1,2)  1  (k+1,1) (2)
  (k+1,1) (1)

d X d K X z X z
2H
 
1  (k+1,1) (2)
  (k+1,1) (k)
  (k) (1)

= d K X z X x + x z .
2H
By d (AB) d (A) |B| + |A| d (B) [4] we have
 
 (k+1,2)  1  (k+1,1) (2)
  (k+1,1) (k)
  (k) 
(1) 2
d X d K X z X x + x z
2H
 
1 K  (k+1,1) (2)
  (k+1,1) (k)
  (k) 
(1) 2
d X z X x + x z .
2 H
Again, we use d (AB) d (A) |B| + |A| d (B) and we have
 
 (k+1,2)  1 K  (k+1,1) (2)
   (k+1,1) (k)
  (k)  
(1) 2 
d X d X z  X x + x z 
2 H
   
 K  (k+1,1)   (k+1,1)   (k)  
(1) 2
+  X (2) 
z d X x (k)
+ x z
H
    
1  (k+1,1)     (k+1,1)
(2)  K 
 K
d X z 
+ X z d (2) 
2  H  H
 (k+1,1)    2
X x(k)  + x(k) z (1)  +
  
 K   (k+1,1)     (k)   (k)  
(k) 2
  X z (2) 
d X X (k)
+ X X .
H 
An improved parameter regula falsi method 1353

   
K K 
By using d   and the fact that d ((X x)n ) 2 (d (X))n , we have
H H

    
 (k+1,2)  1   (k+1,1)   K   (k+1,1)  K 
d X 2 d X   + X z   
(2)
2 H  H
 
 (k+1,1)   2  K   
X x(k)  + x(k) z (1)  +   X (k+1,1) z (2) 
H
     (k)   (k)   (k) 2 
(k)
d d X ,d X + d X ,d X .

By using the same properties, we have

    
 (k+1,2)  1   (k+1,1)   K   (k+1,1)  K 
d X 2 d X   + X z   
(2)
2 H  H
 
 (k+1,1)   2  K   
X x(k)  + x(k) z (1)  +   X (k+1,1) z (2) 
H
   (k)   (k) 2 
d 2d X , 2d X .

Then, we use property d (An ) n |A|n1 d (A) and what we have is

     
 (k+1,2)  1  K   (k+1,1)   K   (k+1,1) 
d X  
2 d X 
+  X   z (2) 
2 H H
 
 (k+1,1)   (k) 2  K   (k+1,1) 
X x  + x z  +   X
(k) (1)
z (2) 
H
   (k)   (k)   (k) 
2  2d X , 2d X  4d X .

       
By using X (k) x(k)  d X (k) and x(k) z (1)  d X (k) , we then have

     
 (k+1,2)  1  K   (k+1,1)   K   (k+1,1) 
d X 2   d X +   X X (k+1,1) 
2 H H
 (k)   
X X (k)  + d X (k) 2 +
  
 K   (k+1,1)    (k) 2
  X X (k+1,1) 
16 d X
H 
     
1  K   (k+1,1)   K   (k+1,1) 

2 d X 
+  d X 
2 H H
   
  (k)   (k) 2  K   (k+1,1)    (k) 2
d X +d X 
+  d X  16 d X
H
1354 N. A. Bakar, M. Monsi and N. Hassan

Simplifying the inequality, we have


   
 (k+1,2)  1  K   (k+1,1)    (k) 2
d X = 3   d X 2d X +
2 H
  
 K   (k+1,1)    (k) 2
 
16   d X d X
H
   
1  K   (k+1,1)   (k) 2
=  
12   d X d X +
2 H
  
 K   (k+1,1)    (k) 2
 
16   d X d X
H
   
1  K   (k+1,1)   (k) 2
= 28   d X d X
2 H
 
 (k+1,2)  K     2
d X = 14   d X (k+1,1) d X (k)
H

Therefore,
 
 (k+1,2)  K     2
d X 14   d X (k+1,1) d X (k)
H
    
 K  1  K   (k)   (k1)   (k) 2

= 14      d X d X d X
H 4 H 
 2
 (k+1,2)  7  K   (k) 3  (k1) 
d X d X d X
2 H 

The solution to this simple recursion is the relation


 (i1)  i
  7  K   (k) 2i1  (k1) 
d X (k+1,i)   d X d X
2 H 
   2i1  (k1) 
d X (k+1,i) i d X (k) d X ,

with
 (i1)  i
7 K 
i =   , i = 1, 2, . . . , p.
2 H 

Using X (k+1) = X (k+1,p) , we get


   2p1  (k1)   
d X (k+1) p d X (k) d X p 0 .

Next, we let X (k+1,p,1) = X (k+1,p) = X (k+1) and we have


   
d X (k+1,p,1) d X (k) .
An improved parameter regula falsi method 1355

Then, it follows that


   
d X (k+1,p,2) d X (k+1,p,1)


 (k+1,p,2)  z (l) z (l1)  (l) 
d X d z (l) f z +
f (z ) f (z
(l) (l1) )

1  (k+1,p,1) (l)
  (k+1,p,1) (l1)

K X z X z .
2
By using the same approach as above, then we will get the following result
 
 (k+1,p,2)  K     
d X 4   d X (k+1,p,1) d X (k)
H
   p1  p 
K  7  K   (k) 2p1  (k1)   (k) 
= 4     d X
H  d X d X
H 2
 p1  p+1
7 K     
= 4   d X (k) 2p1+1 d X (k1) .
2  H 

By using the same mathematical induction on l, then we have


 p1  p+(l1)
 (k+1,p,l) 7 K   2p+(l2)  (k1) 
d X 4 (l1)   d X (k) d X
2  H 

or we can write it as
   (k) 2p+(l2)  (k1) 
d X (k+1,p,l) l1
p d X d X

where
 p1  p+(l1)
7 K 
pl1 =4 (l1)   , 2 l p.
2 H 

For l = p, then X (k+1) = X (k+1,p,p) and we get the relation


   (k) 2p+(l2)  (k1) 
d X (k+1) l1
p d X d X (8)

where
 p1  p+(l1)
7 K 
pl1 =4 (l1)   , 2 l p.
2 H 

Therefore, based on this result and [4] the R-order of convergence of p(l)-RF
method is
  
1 2
OR ((p(l) RF) , ) (2p + (l 2)) + (2p + (l 2)) + 4
2
1356 N. A. Bakar, M. Monsi and N. Hassan

where p 2, 2 l p.
If we compare the R-order of convergence of p(l)-RF method and p-RF
method, so it is clear that p(l)-RF method is more eective. Let p = 5, then
lets substitute the value of p into (3) rst. Hence,
1  
OR ((p RF) , ) p + p2 + 4
2
1 
= 5 + 52 + 4
2
1 
= 5 + 29
2
1 
OR ((p RF) , ) 5 + 29 .
2
Next, substituting p = 5 into (3) gives
  
1 2
OR ((p(l) RF) , ) (2p + (l 2)) + (2p + (l 2)) + 4
2
  
1 2
= (2(5) + (l 2)) + (2(5) + (l 2)) + 4
2
  
1 2
= (8 + l) + (8 + l) + 4
2
  
1 2
OR ((p(l) RF) , ) (8 + l) + (8 + l) + 4 .
2

Thus OR (p(l) RF) OR (p RF) for 2 l 5.

5 Numerical Results
We now compare the performance of p(l)-RF method with p-RF method. The
comparisons are based on the CPU (Central Processing Unit) time and number
of iterations which are presented in Table 1 and Table 2 respectively. The
following table contains the test functions used to test the performance of
both methods where the zero of each function contain in the nal interval.

We also present the nal interval results that contain the zeros of the functions
6 and 7 produced by p-RF method and p(l)-RF method  in Table 3 and Table
4 respectively. The stopping criterion used is d X (k+1) <  = 1015 for p = 5
and 2 l 5. The procedure p-RF and p(l)-RF have been implemented using
Matlab R2007a [1] in associate with Intlab [12].
An improved parameter regula falsi method 1357

No. Function, f (x) X (0) Final interval


1 2xe + 1 2e5x [4]
5
[0, 1] [0.13825715505682, 0.13825715505683]
2 x3 2x 5 [13] [2, 3] [2.09455148154232, 2.09455148154233]
3 2
3 x + 4x 10 [6] [1, 2] [1.36523001341409, 1.36523001341410]
x 2
4 e 4x [3] [4, 4.5] [4.30658472822069, 4.30658472822071]
5 xex 0.1 [11] [0, 1.0] [0.09127652716086, 0.09127652716087]
4
6 (x 1)(x + 1) [8] [0.2, 2.0] [0.99999999999999, 1.00000000000001]
9
7 x (x 1) 1 [4] [0.1, 2.2] [1.07576606608683, 1.07576606608684]

Function CPU time (seconds)


p-RF method p(l)-RF method
1 0.022837 0.010817
2 0.020433 0.016827
3 0.015625 0.014423
4 0.025240 0.015625
5 0.028846 0.019231
6 0.034856 0.018029
7 0.032452 0.019231

Table 1: Comparison of CPU time for both methods after converge.

Function Number of iterations


p-RF method p(l)-RF method
k i (i = 2, . . . , p) k i (2 i p)
1 2 5 2 2
2 2 5 2 2
3 2 5 2 2
4 2 5 2 2
5 3 5 2 2
6 3 5 2 2
7 3 5 2 2

Table 2: Comparison of number of iterations for both methods after converge.


1358 N. A. Bakar, M. Monsi and N. Hassan

Figure 1: Comparison of CPU time and number of iterations of both methods.

p-RF Method p(l)-RF Method


[ 0.85358999999999, 1.09498146639512 ] [ 0.85358999999999, 1.09498146639512 ]
[ 0.97528133236619, 1.02316965237787 ] [ 0.97528133236619, 1.02316965237787 ]
[ 0.99513872890716, 1.02316965237787 ] [ 0.99513872890716, 1.02316965237787 ]
[ 0.99513872890716, 1.01935332764408 ] [ 0.99513872890716, 1.00923360327808 ]
[ 0.99513872890716, 1.01547680514940 ] [ 0.99733393674489, 1.00274565275650 ]
[ 0.99513872890716, 1.01194208444040 ] [ 0.99963385731345, 1.00036649028933 ]
[ 0.99513872890716, 1.00907728269631 ] [ 0.99999583298527, 1.00000416704239 ]
[ 0.99787416316147, 1.00202177682740 ] [ 0.99999583298527, 1.00000416704239 ]
[ 0.99989175268693, 1.00202177682740 ] [ 0.99999583298527, 1.00000416704239 ]
[ 0.99993828354072, 1.00006977151653 ] [ 0.99999748724824, 1.00000012407578 ]
[ 0.99999603617088, 1.00000399775332 ] [ 0.99999999682872, 1.00000000001257 ]
[ 0.99999976544888, 1.00000023469400 ] [ 0.99999999848497, 1.00000000001257 ]
[ 0.99999998619637, 1.00000001380423 ] [ 0.99999999887529, 1.00000000001257 ]
[ 0.99999999999969, 1.00000000000029 ] [ 0.99999999999999, 1.00000000000001 ]
[ 0.99999999999996, 1.00000000000001 ]
[ 0.99999999999999, 1.00000000000001 ]
[ 0.99999999999999, 1.00000000000001 ] Already converge
[ 0.99999999999999, 1.00000000000001 ]
[ 0.99999999999999, 1.00000000000001 ]
 
Table 3: The results of p-RF and p(l)-RF methods with f (x) = (x 1) x4 + 1
An improved parameter regula falsi method 1359

p-RF Method p(l)-RF Method


[ 1.00029766383684, 1.12207347842121 ] [ 1.00029766383684, 1.12207347842121 ]
[ 1.06183800081867, 1.08898632512707 ] [ 1.06183800081867, 1.08898632512707 ]
[ 1.07286786740548, 1.08898632512707 ] [ 1.07286786740548, 1.08898632512707 ]
[ 1.07286786740548, 1.08898632512707 ] [ 1.07286786740548, 1.08250938586611 ]
[ 1.07286786740548, 1.08898632512707 ] [ 1.07343530774921, 1.07818318802548 ]
[ 1.07286786740548, 1.08898632512707 ] [ 1.07513541284123, 1.07639744906999 ]
[ 1.07286786740548, 1.08898632512707 ] [ 1.07573926681963, 1.07579286549305 ]
[ 1.07286786740548, 1.08067519964994 ] [ 1.07576212659559, 1.07579286549305 ]
[ 1.07543570827449, 1.08067519964994 ] [ 1.07576212659559, 1.07578604343998 ]
[ 1.07543570827449, 1.07669506897587 ] [ 1.07576212659559, 1.07576632330652 ]
[ 1.07556002323090, 1.07598559639626 ] [ 1.07576605337270, 1.07576606614498 ]
[ 1.07569584953018, 1.07583658668433 ] [ 1.07576606006073, 1.07576606614498 ]
[ 1.07574338687893, 1.07578875214866 ] [ 1.07576606024736, 1.07576606614498 ]
[ 1.07576606254727, 1.07576606935031 ] [ 1.07576606608683, 1.07576606608684 ]
[ 1.07576606394863, 1.07576606616207 ]
[ 1.07576606608683, 1.07576606608684 ]
[ 1.07576606608683, 1.07576606608684 ] Already converge
[ 1.07576606608683, 1.07576606608684 ]
[ 1.07576606608683, 1.07576606608684 ]
 
Table 4: The results of p-RF and p(l)-RF methods with f (x) = x x9 1 1
1360 N. A. Bakar, M. Monsi and N. Hassan

6 Discussion and Conclusion


We can see that in Table 1, the CPU time of p(l)-RF method is lesser than
p-RF method. Table 2 contains the number of iteration k which applied for
the whole steps of both algorithms and inner iteration i in Steps 8-10 of both
methods after the algorithms stopped by using the given convergence criterion.
In certain cases the number of iterations k are the same for both methods, but
the number of inner iterations i for p(l)-RF are lower. This indicate that p(l)-
RF method have reached the stopping criterion and converges to the zero of
the function much faster. For a clearer view on the CPU time and number of
iterations of both methods, we represent the results in a form of bar charts as
in Figure 1(a) and Figure 1(b) respectively.

Furthermore, Table 3 and Table 4 are the computer output of the functions
6 and 7 respectively. It is clear that p(l)-RF method has achieved excellent
performance where the algorithm of p(l)-RF is terminated earlier by using
the same convergence criterion. In fact, we have shown that the R-order of
convergence of p(l)-RF method is greater than does p-RF method or

OR ((p(l) RF ) , ) > OR (p RF, ) .

7 Acknowledgement
We are indebted to Universiti Kebangsaan Malaysia for funding this research
under the grant UKM-GUP-2011-159.

References
[1] A. Gilat, MATLAB: An introduction with applications, Hobuken,
NJ:Wiley, 2008.

[2] E. Hansen et al., Interval forms of Newtons method, Computing, 20


(1978), 153-163.

[3] F. Costabile, M.I. Gualtieri, S.S. Capizzano, An iterative method for the
computation of the solution of nonlinear equations, Calcolo, 36 (1999),
17-34.

[4] G. Alefeld and J. Herzberger, Introduction to Interval Computation, Aca-


demic Press, New York, 1983.
An improved parameter regula falsi method 1361

[5] I. Gargantini, Further applications circular arithmetic: Shoroeder-like al-


gorithms with error bounds for nding zeroes of polynomial, SIAM. J.
Numer. Anal., 15 (1978), 149-154.

[6] J. Kou and Y. Li, An improvement of the Jarrat method, Applied Math-
ematics and Computation,189 (2007), 1816-1821.

[7] K. Wang & X. Feng, New predictor corrector methods of second-order for
solving nonlinear equations, International Journal of Computer Mathe-
matics, Vol.88,2 (2011), 296-313.

[8] N.A. Abd Rahmin, M. Monsi, M.A. Hassan, F. Ismail, A modication of


inclusion of a zero of a function using interval method, Malaysian Journal
of Mathematical Sciences, 3 (2009), 67-82.

[9] N.A. Bakar, M. Monsi, M.A. Hassan, W.J. Leong, On The Modication
of The p-RF Method, Applied Mathematical Sciences, 5 (2011), 2571 -
2579.

[10] P.K. Parida, & D.K. Gupta, A cubically convergent iteration method for
multiple roots of f (x) = 0, International Journal of Computer Mathemat-
ics, Vol.87,3 (2010), 877-884.

[11] S.A. Hoda Ibrahim, An improved exponential regula falsi methods with
cubic convergence for solving nonlinear equations, J. Appl. Math. & In-
formatics, 28 (2010), 1467-1476.

[12] S.M. Rump, INTLAB - INTerval LABoratory:In Tibor Csendes, Devel-


opments in Reliable Computing, Kluwer Academic Publishers, Dordrecht,
(1999), 77-104. http://www.ti3.tu-harburg.de/rump/.

[13] X. Wu and D. Fu, New high-order convergence iteration methods with-


out employing derivatives for solving nonlineat equations, Computers &
Mathematics with Applications, 41 (2001), 480-495.

Received: September, 2011